Message157935
Rik, I don't follow your argument on not changing discover. Currently, if code calls discover with pattern=None, there will be an exception. So there cannot be any working code out there which passes pattern=None. Therefore, it should be all right for the implementation to detect the None and replace it by a default. No currently working code should be affected, and new code could be shorter that way. The pattern still wouldn't be stored inside the loader, so the doc there still holds. Only the fallback for None would be stored.
By the way, what's the rationale behind passing the pattern to the load_tests function? Shouldn't each package decide which of its files constitute the test suite, independent of what the parent module used as a filter? |
|
Date |
User |
Action |
Args |
2012-04-10 05:42:26 | gagern | set | recipients:
+ gagern, ezio.melotti, eric.araujo, michael.foord, rik.poggi |
2012-04-10 05:42:26 | gagern | set | messageid: <1334036546.49.0.741996683976.issue11218@psf.upfronthosting.co.za> |
2012-04-10 05:42:25 | gagern | link | issue11218 messages |
2012-04-10 05:42:25 | gagern | create | |
|